From 0 to 100 times: AI investment practice that ordinary people can learn

Author: Changan, Amelia I Biteye Content Team

What? Someone used AI to trade cryptocurrencies and made 480 times in 8 days?

In the past, financial markets were a hunting ground of information asymmetry. Retail investors lacked capital, but even more so, they lacked the computing power to process massive amounts of data, the energy to stay awake 24 hours, and the discipline to combat human greed.

Now, AI has become that "Archimedean lever." As long as your logic is correct, AI is the tenfold leverage that helps you pry open wealth.

Here are four major financial experiments featuring four hardcore AI practical applications.

Perpetual Contracts: From 100 to over 100,000, the power of rule execution

📌 Case Review

Lana@lanaaielsa (XHunt ranking: 12280) had Claude help him write a script: to scrape the highest traffic posts from Binance Square, filter out bot accounts, and identify the most volatile assets on the gainers list—buy in and set stop-loss. The entire process was executed fully automatically by AI. In 8 days, the account grew from 100 USDT to 48,000 USDT. As of April 14, Lana's Binance real account profit had reached 146,000 USD.

Two concurrent experiments (Nof1.ai and Aster) also confirmed: AI systematically outperforms humans in risk control—no emotional over-leveraging, no panic selling, no greedy chasing. Absolute returns may not be top-notch, but the advantage lies in not making big mistakes and not incurring large losses.

🧠 Methodology Summary

1️⃣ Information Filtering

He had Claude write a script to automatically scrape the highest volume posts and the most discussed cryptocurrencies in Binance Square daily. The square is a gathering place for retail investor information; his logic is that before the big players pump the market, there must be fish, and the square's popularity is an early signal for retail investors to enter.

2️⃣ Signal Identification

Based on the square data, he overlaid the gainers list. He was not looking for the coins that rose the most, but rather the coins with the largest volatility: high volatility means capital is moving, and where there is movement, there are trading opportunities. He also observed assets with significant changes in open interest (OI) within 48 hours but no immediate price reaction; these coins often signal that capital is positioning itself in advance.

3️⃣ Style Distillation

He distilled his Twitter style and the content of KOLs like the market leader into the AI, allowing it to learn their posting logic and coin selection thought processes, aiding in the judgment of market sentiment and hot directions.

When he asked AI why it chose a certain coin, AI responded that it was because the highest traffic post was retweeted by CZ, and that post mentioned the book "Binance Life," which was the most discussed event in the past three days.

4️⃣ Rule Execution

After buying, he set a stop-loss, posted in the square, and took screenshots of profits to maintain interest. The rules were designed by him: initially setting a 20% stop-loss, later changing it to stop-loss at 200 USDT regardless of position size, only pursuing one direction and not taking counter positions, with AI responsible for execution.

🔧 Tool Summary

• Claude: Writing scripts (data scraping, posting, setting stop-loss orders).

• Binance Square: Data source (post volume, discussion volume, gainers list).

💡 Biteye Reflection

• In the entire process, what AI does is: write scripts, scrape data, and post. The trading strategy is his; AI merely automates these tasks. In the contract market, executing rules more steadily than others is an advantage in itself.

• Action Strategy: First, write down your stop-loss rules: when to exit based on losses, which direction to pursue, and not to chase counter positions. The framework can borrow from Lana's, but the strategy must be your own.

Prediction Markets: Arbitrage + Information Asymmetry + Automation

Prediction markets (like Polymarket) have simple rules: each question is Yes/No, with prices from 0-1 representing probabilities.

🧠 Methodology Summary

The community uses AI to profit in three directions:

• Arbitrage: In the Neg Risk market, using AI scripts to regularly scan the total Bid prices of all Neg Risk markets, automatically filtering opportunities greater than 1, executing Split + sell.

• Reducing Information Asymmetry: Using the open-source project worldmonitor to aggregate over 435 global news sources, covering 15 categories including military, economy, geopolitics, disasters, and finance. AI synthesizes these information streams into briefings in real-time and executes cross-signal correlation analysis. It detects leading signals for geopolitical events in advance.

• Strategy Automation: Describing one's trading judgment framework to AI in natural language, allowing AI to convert it into an executable script. The script automatically monitors trigger conditions, calculates position sizes, and executes orders according to strategy logic.

🔧 Tool Summary

• Polymarket: Mainstream prediction market platform.

• worldmonitor (GitHub 44k+ stars): Real-time global intelligence dashboard, aggregating 435+ news sources, AI-generated briefings + cross-signal correlation analysis.

• Claude / GPT: Converting natural language strategies into scripts.

💡 Biteye Reflection

1. Arbitrage requires technical foundation; information asymmetry is more suitable for beginners: first bookmark worldmonitor, spend 10 minutes daily reading briefings, and find a small position to test an event you can judge.

2. The key to information asymmetry arbitrage is "leading signals": do not chase news, but rather track changes in non-mainstream data sources before the news occurs.

3. Strategy automation is an advanced form: once you have a stable and profitable manual framework, consider using AI to turn it into a program.

Cryptocurrency Spot: K-Line Large Model, Turning Charts into Probabilities

In addition to event and narrative-driven approaches, AI is also undergoing revolutionary changes in the technical aspects of spot trading.

📌 Case Review

The trending GitHub project Kronos tokenizes OHLCV data and pre-trains using autoregressive Transformers on historical data from multiple markets. Retail investors no longer need to memorize dozens of patterns—models directly provide the probability of BTC/USDT rising in the next 24 hours, the probability of volatility amplification, and Monte Carlo simulation paths. The project allows fine-tuning, enabling users to continue training with their own asset data.

🧠 Methodology Summary

The reason large language models can understand text is that they learn the statistical relationships between words from massive amounts of text. Kronos applies the same logic to K-lines: first, using a specially designed tokenizer to convert OHLCV data into discrete token sequences, then pre-training an autoregressive Transformer on these tokens.

The training data covers historical data from 45 global exchanges. After the project launched, GitHub stars quickly surpassed 11,000, with over 2,400 forks.

In the past, retail investors had to memorize dozens of patterns and repeatedly overlay indicators, ultimately relying on personal experience. Now the path has completely changed; you do not need to practice reading charts painstakingly; you can leverage a model pre-trained on massive multi-market data to extract signals.

The project also provides a complete fine-tuning process; if you have historical data for specific assets, you can continue training on the base model to make it better understand your trading targets. It also offers a live demo for BTC/USDT's future 24 hours, accessible to anyone to view real-time prediction results, with the model providing the probability of rising within 24 hours, the probability of volatility amplification, and a forecast chart for the next 24 hours: blue represents historical prices, and the orange line is the average predicted path from multiple Monte Carlo simulations.

💡 Biteye Perspective

1. No need to practice technical analysis painstakingly: in the past, you had to remember dozens of patterns and stack a bunch of indicators; now you can directly use model outputs as references.

2. Observe first, then trade: check Kronos's live demo once a day, comparing model predictions with actual trends to cultivate "probability thinking."

U.S. Stocks: AI Agent Captures Geopolitical Crises, Profiting from Expectation Gaps

📌 Case Review

XinGPT@xingpt (XHunt ranking: 898) built a geopolitical crisis monitoring system using an AI Agent. At that time, the market's focus was on the Strait of Hormuz, with significant noise. His Agent directly monitored first-hand data sources: JMIC ship traffic, Iranian official news agencies, maritime intelligence sources, capturing core indicators every 6 hours—"the actual number of ships passing through the strait." This number dropped from 153 ships/day to single digits, indicating that the situation had not genuinely eased. Based on this, he held an oil ETF since March 7, enduring the pullback until Brent crude rose from 87 USD to over 100 USD.

🧠 Methodology Summary

1. Information Source Planning: First, identify high-quality, low-noise first-hand data sources (official agencies, maritime data, local news agencies), rather than letting AI blindly crawl the entire web.

2. Core Indicator Capture + Noise Filtering: Focus on one most honest indicator (ship traffic), setting up a Flash Alert mechanism to ignore market noise.

3. Decision Framework Automation: Write a separate "Investment Decision Skill" for the Agent, automatically generating daily reports containing signals and position suggestions every morning.

🔧 Tool Summary

• AI Agent (can be built on AutoGPT, LangChain, etc.).

• Data Sources: JMIC (Joint Maritime Information Center), Tasnim/Fars (Iranian news agency), Windward (maritime intelligence).

• Investment Decision Skill: Custom scripts/prompts that automatically generate daily briefings.

💡 Biteye Perspective

1. Framework is more important than tools: first, choose a sector you can track long-term (AI, semiconductors, energy), then find a reliable investment bank research report framework, and finally use Claude to help you build daily briefings.

2. Focus on one core indicator: do not attempt to monitor all variables. Find that indicator at the "ship traffic" level that best reflects the real situation.

3. Making money in U.S. stocks relies on information processing speed and expectation gaps: retail investors find it challenging to timely and comprehensively digest financial reports, macro data, geopolitical events, and industry intelligence, but AI can process massive amounts of information in minutes, identifying opportunities that the market has not fully priced.

In Conclusion

In the past, financial markets were far from ordinary people, with information asymmetry, insufficient capital, unaffordable tools, and a long time to accumulate experience.

Now, AI has almost erased the once insurmountable technical barriers; you only need to communicate your logic to AI in natural language, and it can help you write scripts, scrape data, analyze, and execute.

Lana made 480 times in 8 days, XinGPT steadily profited during macro crises, and ordinary people can use models like Kronos to turn K-lines into probability predictions. These tasks, once only achievable by professional teams, can now be done by novices sitting at home with a computer.

AI does not bring the illusion of "everyone can get rich," but rather true technical equality: equality in information access, analytical capability, execution efficiency, and decision-making systems.

To start from here, you can implement these three steps:

1. Choose a market you are most interested in and find 2-3 KOLs you can track long-term.

2. Distill their recent content into Skills, allowing AI to extract their judgment logic.

3. Describe your strategy clearly in natural language, letting AI help you write an automated script.

The first pot of gold never belongs to the richest, but to those who can leverage AI and systematize their judgment frameworks.